OpenAI publishes 722 maths papers claiming progress on 372 problems
OpenAI has released 722 mathematical manuscripts on GitHub, claiming progress on 372 major unsolved problems. The work used an unreleased ChatGPT model and was reviewed by OpenAI's independent Advisory Group on Mathematics and Artificial Intelligence. The release is expected to intensify scepticism in the mathematics community, which has been wary of AI-generated results following previous controversial claims.
The claimed breakthroughs include a solution to the four-dimensional Kakeya conjecture and progress towards the Riemann hypothesis, with each average result requiring approximately three hours of ChatGPT Pro use. However, leading mathematicians including MIT researcher Andrew Sutherland have demanded that OpenAI release the underlying model to allow independent verification, arguing that AI claims in mathematics require peer scrutiny before acceptance. OpenAI did not release specific compute times or the exact prompts used, despite guidance from its advisory group.
- OpenAI posts 722 manuscripts claiming solutions to 372 major maths problems
- Key results include Kakeya conjecture solution and Riemann hypothesis progress
- Mathematicians remain sceptical and demand independent verification before accepting results
New here? Start with this
OpenAI is an artificial intelligence company that has released a large collection of mathematical papers. The company claims these papers show progress on major mathematical problems that have never been solved. To make sense of this news, it helps to understand that mathematics has many famous unsolved problems that experts have worked on for decades or even centuries.
Artificial intelligence systems have become increasingly powerful in recent years, and companies are now exploring whether they can tackle these difficult mathematical questions. However, mathematicians have grown sceptical of claims about AI breakthroughs following some disputed results in previous years. In mathematics, a proposed solution only gains acceptance once other experts have carefully examined it to check for errors.
The question now is whether mathematical work produced by AI should be treated the same way as work by human mathematicians, and what standards should apply for proving that AI has genuinely solved a problem. This matters because if AI systems can truly solve major mathematical puzzles, it could change the field fundamentally, but only if those solutions stand up to expert scrutiny.
Both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
OpenAI's release represents a meaningful contribution to mathematical progress that warrants community engagement and scrutiny. By publishing their work on GitHub with independent advisory review, they are facilitating transparency and allowing mathematicians to examine, verify, and build upon the findings themselves. Releasing working papers before full model access enables broader participation and feedback, accelerating scientific progress rather than gatekeeping results behind proprietary walls. The mathematical arguments can be assessed on their merits independently of the model's internal workings.
The case against
Rigorous peer review and complete reproducibility are foundational standards that protect mathematical knowledge's integrity and cannot be compromised even for AI contributions. Without OpenAI disclosing the exact model, prompts, and computational parameters, genuine independent verification—essential before accepting extraordinary claims about decades-old problems—is impossible. The mathematics community's caution is justified given prior overstated AI claims, and publishing through GitHub rather than traditional peer-review venues circumvents the careful scrutiny such breakthroughs demand. Full methodological transparency is not optional when claiming significant mathematical progress.
AI Business Companies Technology
Read the full article at the source →
Originally published by Engadget as “OpenAI just posted hundreds more results on major math problems”.